A Korner-Marton approach to low complexity video encoding

We consider a distributed video coding paradigm. Contrary to conventional paradigms, which perform motion estimation at the encoder, the distributed paradigm encodes each frame individually and jointly decodes subsequent frames. Therefore, encoding complexity is significantly reduced and transferred to the decoder. Our scheme is based on the Korner-Marton setup, which demonstrates how the difference between two signals can be compressed in a distributed manner using syndromes of a linear code. This approach is different from prior propositions, based on the Wyner-Ziv Theorem (source coding with side information). In the Predictive Korner-Marton (PKM) scheme, the encoder transmits only the least significant bits (LSBs) of the quantized pixels. The decoder then compensates for the missing MSB information by motion estimation based on the data's temporal structure. Our results show potential to this framework, both in terms of motion estimation accuracy and encoding complexity reduction, although the overall compression performance is still inferior to that of conventional video codecs.